JOURNAL OF ALGEBRA 146, 267-277 (1992) On the Representation Theory of Modular Hecke Algebras* PETER SIN Depariment of Mathematics, Llniversity of Florida, Gainesville, Florida Communicated by Walter Feit Received February 21, 1990 The Cartan invariants and decomposition numbers of the endomorphism algebra of the permutation module on the cosets of a Sylow p-subgroup of a finite group of Lie type in characteristic p are computed. 6 1992 Academic Press. Inc. INTRODUCTION Let G be a finite group with a split &V-pair of characteristic p. We shall be interested in (right) modules over the group algebra over a complete discrete valuation ring 0 of characteristic zero, its field of quotients K and its residual field k. We shall assume that k has characteristic p, and that K and k are splitting fields for all subgroups of G. For any O-lattice& we use the notations L ,:=L OO K and L:=L Ofi k. Let Y=ind,.(O) be the permutation UC-lattice on the right cosets of the Sylow p-subgroup U of G, and E = End,,(Y). Then Y, = ind,.(K) and P= ind,.(k). Moreover, E is an Co-formof End&Y,) and E= End& P). In this paper we shall com- pute the Cartan invariants of ,?, and the decomposition numbers of the tri- ple (E, E,, E) in terms of the Weyl group associated to G. Some earlier work in this direction has been done by Norton [12], who computed the Cartan matrix and blocks of End,,(ind,,(k)), which is an algebra direct summand of & and by Carter [4], who described the decomposition numbers in type A,, again for the subalgebras obtained by replacing the subgroup U by its normaliser B. Recently, Cabanes [2] has determined the extension groups and blocks of E. He has also given a different description of the Cartan invariants. The approach in all of these papers is via the well- known description of the algebra by generators and relations [2, Sect. 21. We shall instead exploit the fact that i? is the endomorphism ring of a * This research was supported by NSF Grant DMS 8802740. 267 0@21-8693/92 $3.00 Copyright 6’ 1992 by Academtc Press. Inc All rights of reproduction m any form reserved 268 PETER SIN module about which much is known [S, 14, 171. The operation of E on the left of P defines an (E, kG)-bimodule structure on Y. It is well known [S, Prop. 6.31 that the restrictions of the functors Hom,,( Y, -) and (-) BE P are inverse k-equivalences between the full subcategory add P of finitely generated right kG-modules with indecomposable direct summands isomorphic to those of P and the category of finitely generated projective right E-modules. (Corresponding equivalences also hold over 0 and K.) It is also known [17] that i? is a Frobenius algebra. It follows that the Cartan invariants are nothing other than the intertwining numbers between the various indecomposable direct summands of Y. Since these summands are trivial source modules, the calculation of their intertwining numbers is reduced to that of the inner products of the ordinary characters afforded by lifts to characteristic zero of these modules. We shall compute these using an induction formula, proved by N. Tinberg [ 171 for characters and extended to the Green ring in [ 143. There, we computed these numbers for the submodule ind,,(k), and observed a duality phenomenon. This case had previously been calculated in [ 121. In Section 4, we prove that at least for the summands of the permutation module on the cosets of the subgroup B, this duality is the restriction of the one studied by Curtis [6] and Alvis [l]. This follows from the formula we shall obtain for the decomposition numbers of the triple (E, E,, E). This formula is the same as one which appears in work of Stanley [lS], and so the decomposition numbers have the combinatorial description given to this formula there. In particular, for type A,, our calculation of the decomposition numbers explains why the same numbers appear in [ 15,4]. 1. PRELIMINARIES We recall some notation and basic facts (cf. [3]); thus we have sub- groups B, N, H, U, with UE SylJG), B = UH, H being abelian of order prime to p, H = B n N a N, and a Weyl group W= N/H= (wi 1 iG R), where R is the set of fundamental roots corresponding to B in the root system @ on which W operates. For each subset S of @ we denote by W, the subgroup of W generated by the elements w,, s E S. For each subset JE R, we have a standard parabolic subgroups N, of N and GJ= (B, NJ) of G such that G, = B and G, = G, and a parabolic subgroup W,= (wi 1iE J). It follows from the remarks after [3, p. 60, Corollary 2.6.21 that it is possible to pick coset representatives n,, w E W so that each lies in the kernel of every character x: G, + k* of every MODULAR HECKE ALGEBRAS 269 standard parabolic subgroup to which it belongs. We shall fix such a choice throughout. Let wR be the longest element of W and let U, be the sub- group U n UnwRnn.If Y is a root system with base A and Weyl group W,, we define for JE L E A, X(Y),= (WE W, I wp’(J)>O), Z(Y),,= {WE W, I w-‘(J)>O, wp’(L\J)<O}. The set X( Vu), is a set of right coset representatives of W,, in W, and the set X(Y), n X( !F);’ is a set of ( W,,, W,,) double coset representatives in W, [3, Proposition 2.7.31. The set {n, 1w E A’(@),) is a set of right coset representatives of G, in G, and for J, K E R, the set is a set of representatives for (G,, GK) double cosets in G and for (NJ, Nk) in N [3, Proposition 2.8.11. We shall identify the group of K-characters and the group of k-characters of H via the natural map 0 -+ k, and denote the group by fi. No confusion will arise since we shall be interested in the action of W on these groups by A”‘(h)=I(n,hn;‘), where A is a (K- or) k-character of H, and the map above is clearly a W-map. Let X, be the root subgroup corresponding to the root tx (see [3, p. SO]), and for i E R, denote by Hi the subgroup H n (X,, Xi). For II E Hom(H, k*), we define the set M(A)= {iER I II,z= l}. Let Wj~ be the stabilizer of A and W(l) the (normal) subgroup of W, generated by its reflections. Then W(n) is the Weyl group of a root system Qn E @ with base R, containing M(1) (but the inclusion R, 5 R need not hold). Moreover, W, is the semidirect product of W(2) with the subgroup C, := (w E W, I w(R,) = R,}. These facts are due to Kilmoyer [lo] and Howlett-Kilmoyer [9] and can be found in [3, Proposition 10.6.31. Curtis [S] has shown that the module Y has the following decomposi- tion into pairwise nonisomorphic indecomposable direct summands: y= @ p(x, J). (2T.J) Here the indices (x, J) run through all pairs consisting of an element x of I? and a subset J of M(X). Let L, be a k&module affording x. Then [S], L, extends (uniquely) to G,,,M(XJ,and we have the following formula in the Green ring of kG [ 14, Theorem 21: F((x, J) = 1 (- l)ix’J’ ind,,,,(L,). (*) Jz Kc M(X) 270 PETERSIN Finally, we shall use “( , )” for the usual inner product of (ordinary) characters, and if CIis a character of a subgroup D of a group A, we shall write U$ for the induced character. 2. COMPUTATION OF THE CARTAN INVARIANTS We denote the projective indecomposable E-modules by - - -h J) := Homd K Y(x, J)). Thus, we have E= @ E(x, J), (x.4 where the indices run over all admissible pairs as above. The Cartan invariants are given by c (A,~),(~.~)= dimkWomdh Jh .&A K))). Suppose L and x are elements of I? with x = I’” for some element w E W. Then clearly W(x) = W(A)w and so there is an element w’ E W(x) such that w’w(R,)= R,. For each subset KG R, we let K’ := w’w(K). Then K’ depends on the choice of w but different choices give sets which are conjugate by an element of C,. The aim of this section is to prove: THEOREM 1. (a) If x and A are not W-conjugate then all of the C (A,~),(~,~)are zero. (b) rfx and I are W-conjugate, then The following result is well known and stems from an idea of Steinberg [ 16, Lemma 2.23. LEMMA 2. For 1, x E fi, K E M(l), and JG M(X), we have Cx& C,) = CiGJYG,). Proof We may take the same set of double coset representatives on both sides (cf. Section 1) and then by the Mackey formula, both sides are equal to the number of these double coset representatives which map x to R. MODULAR HECKEALGEBRAS 271 LEMMA 3. Suppose x and 2 are W-conjugate. Then for JG M(X) and K E M(A) we have in the notation of Section 1, By Clifford theory, the last inner product is equal to where N, is the stabiliser of x in N, which is clearly the inverse image in N of W,.